Recent studies revealed that microcircuits in the brain consist of multiple heterogeneous cell types that interact with each other forming highly specific connectomes that evolve over time and with experience. Interestingly, earlier modeling studies have shown that the microcircuits experimentally observed can be directly related to perception and high-level cognitive functions (e.g., decision-making). Further, deep learning can create artificial microcircuits capable of performing complex functions. Thus, it is imperative that we study microcircuits' structure and functions to better understand the operating principles of the brain and artificial neural networks. To this end, interdisciplinary approaches are essential, including integrating the fields of neuroscience and deep learning.
To study microcircuits' functions, we can employ either an inductive or deductive approach. In an inductive approach, we can first analyze neural data to identify microcircuits' structure and then use computational models to simulate their functions. In a deductive approach, we can first determine the functional building blocks (FBB) of cognitive functions and build microcircuits that can support FBB. Then, we can test predictions of constructed microcircuits against experimental data. It should be noted that the deductive approach is rarely explored partially because our understanding of FBB remains limited, but deep learning (DL), which uses artificial neural systems, can provide valuable insights into FBB. Thus, developing analysis tools of DL models is necessary for such a deductive approach.
This Research Topic seeks any research that can advance both approaches mentioned above, but given that the deductive approach is less explored, we especially welcome contributions addressing this deductive approach. We also welcome review papers discussing the current understanding of microcircuits in biological and artificial neural systems.
All contributions related to functions of microcircuits of neural systems are welcome including, but not limited to, contributions addressing the following themes:
1. Novel analysis methods that can identify cell types and connectomes between them from experimental data (e.g., high density physiological recording)
2. Computational modeling studies that can provide new insights into the functions of neural circuits experimentally observed
3. Open-source datasets or toolkits that can be used for biologically plausible computational models and a new set of analyses
4. Novel theories on cell types' functions and their potential roles in deep learning
5. Algorithms that can map between physiological recordings and other imaging techniques (e.g., fMRI or calcium imaging)
6. Novel methods to study neural representations in either biological or artificial neural systems
7. Studies proposing 1) neuro-inspired computing/learning algorithms, 2) neuroscience-inspired analysis tools of DL, 3) DL-inspired analysis of the brain, and 4) DL-inspired computational modeling of the brain
Recent studies revealed that microcircuits in the brain consist of multiple heterogeneous cell types that interact with each other forming highly specific connectomes that evolve over time and with experience. Interestingly, earlier modeling studies have shown that the microcircuits experimentally observed can be directly related to perception and high-level cognitive functions (e.g., decision-making). Further, deep learning can create artificial microcircuits capable of performing complex functions. Thus, it is imperative that we study microcircuits' structure and functions to better understand the operating principles of the brain and artificial neural networks. To this end, interdisciplinary approaches are essential, including integrating the fields of neuroscience and deep learning.
To study microcircuits' functions, we can employ either an inductive or deductive approach. In an inductive approach, we can first analyze neural data to identify microcircuits' structure and then use computational models to simulate their functions. In a deductive approach, we can first determine the functional building blocks (FBB) of cognitive functions and build microcircuits that can support FBB. Then, we can test predictions of constructed microcircuits against experimental data. It should be noted that the deductive approach is rarely explored partially because our understanding of FBB remains limited, but deep learning (DL), which uses artificial neural systems, can provide valuable insights into FBB. Thus, developing analysis tools of DL models is necessary for such a deductive approach.
This Research Topic seeks any research that can advance both approaches mentioned above, but given that the deductive approach is less explored, we especially welcome contributions addressing this deductive approach. We also welcome review papers discussing the current understanding of microcircuits in biological and artificial neural systems.
All contributions related to functions of microcircuits of neural systems are welcome including, but not limited to, contributions addressing the following themes:
1. Novel analysis methods that can identify cell types and connectomes between them from experimental data (e.g., high density physiological recording)
2. Computational modeling studies that can provide new insights into the functions of neural circuits experimentally observed
3. Open-source datasets or toolkits that can be used for biologically plausible computational models and a new set of analyses
4. Novel theories on cell types' functions and their potential roles in deep learning
5. Algorithms that can map between physiological recordings and other imaging techniques (e.g., fMRI or calcium imaging)
6. Novel methods to study neural representations in either biological or artificial neural systems
7. Studies proposing 1) neuro-inspired computing/learning algorithms, 2) neuroscience-inspired analysis tools of DL, 3) DL-inspired analysis of the brain, and 4) DL-inspired computational modeling of the brain